首页> 外文会议>Parallel Processing Workshops, 2009. ICPPW '09 >The Rendezvous Mechanism for the Multi-core AMBA System
【24h】

The Rendezvous Mechanism for the Multi-core AMBA System

机译:多核AMBA系统的交会机制

获取原文

摘要

The complex architectures of the signal core have the bottlenecks to improve performance. The homogenous and heterogonous multi-core architecture will be the further trends. For the jobȁ9;s cooperation, the synchronous behaviors are needed to keep the task achievements and, more frequently, occur in the multi-core systems. Polling and interrupt usually used as the inter-processor communication mechanisms among multi-cores, and they will cause the heavy bus traffic and more cycle loss for the context switching. In this paper, we have proposed a rendezvous mechanism and implement it at the AHB, called RAHB. In RAHB, the slave organizations will be added in each coreȁ9;s AHB interface. When the data is not ready, the core will be waiting and regarded as a slave device on bus. By this way, the synchronization will be constructed with the load-store instructions among multi-cores without polling mechanisms or interrupt handlers. Because the RAHB reduce the conflicts on the bus and cycle loss, simulation results show that the performance will be improved about the speedup 30% approximately. The core locked hazard had also been considered and avoided by the aging mechanism in this design.
机译:信号核心的复杂架构具有提高性能的瓶颈。异构和异构的多核架构将是进一步的趋势。对于工作9的合作,需要同步的行为来保持任务的完成,并且这种频繁发生在多核系统中。轮询和中断通常用作多核之间的处理器间通信机制,它们将导致繁重的总线流量和上下文切换带来更多的周期损失。在本文中,我们提出了一个会合机制并在称为RAHB的AHB上实现。在RAHB中,将从属组织将添加到每个内核的9个AHB接口中。当数据未准备好时,内核将等待并被视为总线上的从设备。通过这种方式,将使用多核之间的负载存储指令来构建同步,而无需轮询机制或中断处理程序。由于RAHB减少了总线上的冲突和周期损失,因此仿真结果表明,性能大约可提速30%左右而得以改善。此设计中的老化机制也已考虑并避免了岩心锁定的危险。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号